Abstract
It is widely accepted that Pavlovian fear conditioning requires activation of NMDA receptors (NMDARs) in the basolateral amygdala complex (BLA). However, it was recently shown that activation of NMDAR in the BLA is only required for fear conditioning when danger occurs unexpectedly; it is not required for fear conditioning when danger occurs as expected. This study tested the hypothesis that NMDARs in the BLA are engaged for Pavlovian fear conditioning when an animal’s predictions regarding danger are in error. In each experiment, rats (females in Experiment 1 and males in Experiments 2–5) were conditioned to fear one stimulus, S1, when it was paired with foot-shock (S1→shock), and 48 h later, a second stimulus, S2, when it was presented in sequence with the already-conditioned S1 and foot-shock (S2→S1→shock). Conditioning to S2 occurred under a BLA infusion of the NMDAR antagonist, D-AP5 or vehicle. The subsequent tests of freezing to S2 alone and S1 alone revealed that the antagonist had no effect on conditioning to S2 when the shock occurred exactly as predicted by the S1, but disrupted this conditioning when the shock occurred earlier/later than predicted by S1, or at a stronger/weaker intensity. These results imply that errors in the timing or intensity of a predicted foot-shock engage NMDARs in the BLA for Pavlovian fear conditioning. They are discussed in relation to theories which propose a role for prediction error in determining how experiences are organized in memory and how activation of NMDAR in the BLA might contribute to this organization.
SIGNIFICANCE STATEMENT This study is significant in showing that prediction error determines how a new experience is encoded with respect to a past experience and, thereby, whether NMDA receptors (NMDARs) in the basolateral amygdala complex (BLA) encode the new experience. When prediction error is small (e.g., danger occurs as and when expected), the new experience is encoded together with a past experience as part of the same “mental model,” and NMDAR activation in the BLA is not needed for this encoding. By contrast, when prediction error is large (e.g., danger occurs at an unexpected intensity or time), the new experience is encoded separately from the past experience as part of a new mental model, and NMDAR activation in the BLA is needed for this encoding.
Introduction
Over the past 30 years, it has become accepted that Pavlovian fear conditioning requires activation of NMDA receptors (NMDARs) in the basolateral amygdala complex (BLA). This acceptance is largely based on studies in which rats received a BLA infusion of an NMDAR antagonist (e.g., D-AP5) or vehicle and were then exposed to pairings of an initially neutral conditioned stimulus (CS; e.g., a sound) and a brief but aversive unconditioned stimulus (US; e.g., foot-shock). During subsequent testing with the CS alone, rats that received the NMDAR antagonist exhibited less fear than vehicle-infused controls (Gewirtz and Davis, 1997; Rodrigues et al., 2001; Bauer et al., 2002; Parkes and Westbrook, 2010; Holmes et al., 2013). Importantly, this result is independent of the type of stimulus used as the CS (e.g., auditory, visual) and the specific responses used to index the CS-US memory (e.g., freezing, potentiated startle). Hence, the prevailing view is that Pavlovian fear conditioning requires activation of NMDAR in the BLA; and, more generally, that formation of a fear memory always requires activation of NMDAR in the BLA.
However, recent findings show that this is not the case. Williams-Spooner et al. (2022) used a two-stage conditioning protocol in which rats were exposed to pairings of one stimulus (e.g., a sound, labeled S1) and shock in stage 1 (S1-shock) and, 2 d later, to serial presentations of a second stimulus (e.g., a light, labeled S2), the conditioned S1 and shock (S2-S1-shock) in stage 2. Immediately before the latter stage, rats received a BLA infusion of either the NMDAR antagonist, D-AP5 or vehicle alone. The major results of this experiment were from the subsequent drug-free test where Williams-Spooner et al., found that the D-AP5 infusion had no effect on conditioning of the S2: rats that had received the D-AP5 infusion froze just as much as the vehicle-infused controls when tested with S2 alone. This absence of a D-AP5 effect was not because of the prior conditioning of S1 per se, as the drug disrupted conditioning to S2 among rats that had been exposed to S1-shock pairings in stage 1 and S2-[trace]-shock pairings in stage 2 (i.e., S1 omitted in stage 2) and among rats exposed to S1-shock pairings in stage 1 and S2-S1 pairings in stage 2 (i.e., shock omitted in stage 2). Instead, Williams-Spooner and colleagues argued that the involvement of NMDAR in forming the new S2 fear memory reflects the degree to which predictions regarding the US are confirmed or in error. When these predictions are in error (e.g., the US is not predicted by other stimuli that are present, or is predicted but fails to occur), NMDAR in the BLA are engaged for conditioning of the S2. By contrast, when predictions regarding the US are confirmed (i.e., the US is predicted by other stimuli that are present), NMDAR in the BLA are not required for conditioning of the S2.
The present study further examined the role of prediction error in determining whether Pavlovian fear conditioning requires activation of NMDAR in the BLA. It specifically examined whether animals’ predictions about the US include information about its timing and intensity; and, thereby, whether errors in these predictions would engage NMDAR for Pavlovian fear conditioning under circumstances where they are not otherwise required. In each experiment, rats were exposed to S1-shock pairings in stage 1 and then to S2-S1-shock sequences in stage 2 under a BLA infusion of D-AP5 or vehicle. We expected to replicate the finding that conditioning of S2 does not require NMDAR-activation in the BLA when the US occurs as predicted by the already-conditioned S1 (Experiment 1; Williams-Spooner et al., 2022). The question of interest was whether the NMDAR-activation requirement for conditioning to S2 is reinstated when the shock occurs earlier or later than predicted by the S1 (Experiments 2 and 3a, respectively), or at a stronger or weaker intensity than predicted by the S1 (Experiments 4 and 5a, respectively).
Materials and Methods
Subjects
Subjects were 123 experimentally naive, adult, Long–Evans rats (106 males weighing 400–470 g and 17 females weighing 250–300 g) obtained from the Rat Breeding Facility at the University of New South Wales. The initial experiment used 17 female rats to establish the generality of the key finding reported by Williams-Spooner et al. (2022), which used male rats. The remaining experiments used male rats only to facilitate comparisons with the full set of results obtained by Williams-Spooner and colleagues. Rats were housed in plastic tubs (22 cm high × 67 cm long × 40 cm wide) by sex with a maximum of four rats per tub. The tubs were kept in a colony room maintained at 20°C with lights on between 7 A.M. and 7 P.M. Chow and water were available in the tubs across the course of the experiment. Rats were handled for a few minutes each day for a minimum of 5 d before the start of the experiment.
Apparatus
Behavioral procedures were conducted in four identical chambers. Each was 30 cm long × 26 cm wide × 30 cm high and located in its own sound-attenuating and light-attenuating wooden cabinet. The front and rear walls of a chamber were clear Plexiglas, the side walls were aluminum, and the chamber floor was made of stainless-steel rods (7 mm in diameter, spaced 1.8 mm apart). An electric current of a predefined amperage and duration could be delivered through the grid floor to administer shock via a custom-built generator. A tray below the grid floors contained bedding material. LEDs covered by a reflector were mounted on the rear wall of the cabinet behind each chamber and used to deliver a flashing light at 3.5 Hz and ∼8 lux (measured at the center of the chamber). A speaker also mounted on the rear wall of each cabinet was used to deliver a 620-Hz square wave tone at an intensity of 65 dB against a background noise of ∼45 dB. An infrared light mounted on the back wall of each cabinet was used to make the rat visible. A camera, also located on the rear wall, and connected to a computer located in another room in the laboratory, was used to record the behavior of each rat.
Stimuli
The stimuli were the tone and flashing light counterbalanced for their roles as S1 and S2 such that when the tone was S1, the flashing light was S2 and vice versa. The duration of S1 and S2 were 10 s and 30 s, respectively. The US was a 0.8 mA, 0.5 s foot shock, unless otherwise specified.
Surgery
All rats were surgically implanted with bilateral cannulas targeting the BLA. Rats were placed inside an induction chamber and anesthetized with isoflurane (5%) combined with oxygen (1.0 L/min). The anaesthetized rat was then placed on a stereotaxic apparatus. Anesthesia was maintained through continued delivery of isoflurane (1.5–2.5%) and oxygen (0.5–0.7 L/min) via a face mask. Injections of an antibiotic (Benacilin, 0.15 ml/kg) and anti-inflammatory (Carprofen, 1 ml/kg, mixed 1:9 with 0.9% sterile saline) were administered subcutaneously. A local anesthetic (bupivacaine, 0.5%, 0.2 ml) was also administered subcutaneously along the line where the first incision would be made. After the skull was shaved and cleaned, two 26-gauge guide cannulas were implanted into the brain via holes drilled into the skull, with the tips aimed at the BLA in each hemisphere (2.3–2.5 mm posterior to bregma, 4.8–4.9 mm lateral to the midline and 8.1–8.2 mm ventral to bregma). The guide cannulas were then secured in position with dental cement and four jeweler’s screws. A dummy cannula was kept in each guide cannula at all times except during infusions. Immediately after surgery, rats were injected subcutaneously with 0.9% sterile saline (3 ml) and were placed on a heating mat until they had recovered from the effects of the anesthetic. Once recovered, the rats were returned to their home tubs and given a minimum of 7 d to recover from surgery. Rats were handled, weighed, and monitored each day during the recovery period.
Drug infusions
Rats received a bilateral BLA infusion (0.5 μl per hemisphere) of the NMDAR antagonist D-AP5 (10 μg/μl, mixed with artificial CSF (ACSF = 50.7 mm, Sigma-Aldrich) or vehicle (ACSF; Sigma-Aldrich). Before infusion, dummy caps were removed from the guide cannulas and 33-gauge internal cannulas were inserted. The internal cannulas were connected to separate 25 ml Hamilton syringes driven by an infusion pump (Harvard Apparatus), which delivered the drug or vehicle into the BLA at a rate of 0.25 μl/min. Once the infusion had ceased, the internal cannulas were left in place for 2 min to allow for diffusion away from the cannula tip.
Histology
Following behavioral testing, rats were euthanized with an intraperitoneal injection of a lethal dose of sodium pentobarbital, decapitated, and their brains removed and frozen. Brains were sliced on a cryostat into coronal sections of 40-µm thickness, with every second section through the BLA mounted on a glass microscope slide and stained with cresyl violet. Slides were observed under a microscope to confirm the location of cannula tips using the brain atlas of Paxinos and Watson (2007). Rats with one or both cannulas positioned outside the boundaries of the BLA or with extensive damage to the BLA were excluded from the statistical analysis (see Fig. 1).
Cannula placements for all rats in Experiments 1–5 that had both cannulas correctly positioned in the BLA. The placements were verified using Nissl-stained coronal sections and with reference to the atlas of Paxinos and Watson (2007). Each dot represents the most ventral point of a single cannula track. The numbers indicate the sections relative distance (in mm) from bregma (adapted from Paxinos & Watson, 2007).
Scoring and data analysis
Freezing, defined as the absence of all movement apart from that required for breathing (Fanselow, 1980), was the measure of conditioned fear. A time sampling procedure was used in which each rat was scored every 2 s as either freezing or not freezing. The data shown are the number of observations on which each rat was scored as freezing out of the total number of observations. All test data were scored by the experimenter and cross-scored by an experienced observer who was blind to the purposes of the experiment. The Pearson product moment correlation was calculated to assess the reliability of the scores by the two observers. This correlation was >0.9 in this and all subsequent experiments. Any discrepancies between the scores were resolved in favor of those by the naive observer.
The freezing data were analyzed using a mixed model ANOVA with a between-subject factor of infusion type (D-AP5 or vehicle) and a within-subject factor of trial: single trials for conditioning and blocks of two trials for test sessions. The criterion for rejection of the null hypothesis (α) was set at 0.05. For all significant differences, standardized confidence intervals (95% CI) were calculated. Effect sizes were calculated using Cohen’s d (d) for between-subject comparisons and partial η2 (η2p) for within-subject comparisons: 0.8 is considered a large effect for d, and 0.14 a large effect for η2p.
Procedure for Experiment 1
Context exposure
On days 1–2, rats received twice daily exposures to the context, each lasting 20 min, one in the morning and the other approximately 3 h later in the afternoon. This was done to familiarize the rats with the chambers and minimize any neophobic reactions that might obscure the development of freezing across conditioning.
Stage 1: conditioning of S1
On day 3, rats received four S1-shock pairings. Offset of the 10-s S1 co-terminated with a 0.8 mA × 0.5 s foot shock. The first S1-shock pairing occurred 5 min after placement in the context, the intertrial interval (ITI) between pairings (S1 offset to S1 onset) was 5 min, and rats remained in the context for 1 min following the final pairing.
Context extinction
On day 4, rats received two, 20-min exposures to the conditioning chambers in the absence of any scheduled events, one in the morning and the other in the afternoon. This was done to extinguish any context-elicited freezing that might obscure the detection of freezing to the discrete cues across serial-order conditioning.
Stage 2: serial-order conditioning
On day 5, rats received bilateral intra-BLA infusions of either vehicle (n = 7) or the competitive, nonselective NMDAR antagonist, D-AP5 (n = 9). Approximately 10 min later, they were exposed to four S2-S1-shock sequences. Each sequence consisted in the presentation of a 30-s S2, which terminated in the onset of the 10-s S1, which, in turn, co-terminated in the 0.8 mA × 0.5 s foot shock. The first S2-S1-shock sequence occurred 5 min after placement in the chambers, the interval between the sequences was 5 min (S1 offset to S2 onset), and rats remained in the chambers for 1 min following the final S2-S1-shock sequence.
Context extinction
On day 6, rats received two, 20-min exposures to the chambers in the absence of any scheduled events, one in the morning and the other 3 h later in the afternoon to extinguish any context-elicited freezing. Rats received an additional 10-min exposure to the chambers on the morning of day 7 to extinguish any spontaneously recovered context-elicited freezing. Any such freezing could interfere with the detection of freezing elicited by S2 and S1 at test.
Test
On the afternoons of days 7 and 8, rats were tested for freezing to S2 (day 7) and S1 (day 8). Each test consisted in eight presentations of S2 or S1. The interval between presentations was 3 min and the durations of S2 and S1 remained 30 and 10 s, respectively.
Extinction of S2 and its reconditioning
Consistent with the results reported by Williams-Spooner et al. (2022), the test results indicated that the BLA infusion of D-AP5 had no effect on conditioning of S2 (or any additional conditioning of S1). To confirm the efficacy of the BLA D-AP5 infusion, additional training and testing was conducted on days 9–13. On each of days 9 and 10, rats received two extinction sessions of S2, one in the morning and the other in the afternoon. Each of the four extinction sessions consisted in eight presentations of S2 alone. The first presentation occurred 2 min after placement in the chambers, each presentation lasted for 30 s, the interval between the presentations (S2 offset to S2 onset) was 3 min, and rats remained in the chambers for 1 min following the final presentation.
On day 11, rats were infused with D-APV or vehicle into the BLA. 10 min later they received four S2-shock pairings. Each of these pairings involved presentation of the 30-s S2, which co-terminated with the 0.8 mA × 0.5 s foot shock. The first pairing occurred 5 min after placement in the context, the ITI was 5 min, and rats remained in the context for 1 min after the final pairing. The two groups, Vehicle (n = 7) and D-AP5 (n = 9), were composed of equal numbers of rats that been infused with vehicle or the drug before the serial-order conditioning session on day 5. However, eight rats were excluded from this part of the experiment (three from Group Vehicle and five from Group D-AP5) as they continued to exhibit very high levels of freezing to the S2 across the preceding extinction sessions. Thus, the final numbers of subjects in these groups were n = 3 for Group Vehicle and n = 4 for Group D-AP5.
On day 12, rats were exposed to the conditioning chambers for 20 min in the absence of any events and were again exposed for 10 min on the morning of day 13 to extinguish any freezing elicited by the chambers. On the afternoon of day 13, rats were tested with S2 as previously described: the first of the eight 30-s S2 presentations occurred 2 min after placement in the chamber, the ITI was 3 min, and rats remained in the chamber for 1 min after the final S2 presentation.
Procedure for Experiment 2
Context exposure
The details of exposure to the chambers on days 1–2 and of extinction of context-elicited freezing on days 4, 6, and 7 were as described for Experiment 1.
Stage 1: trace conditioning of S1
On day 3, rats received four 10-s presentations of S1. The termination of each presentation was followed 10 s later by a 0.8 mA × 0.5 s foot shock. The first trace-conditioning trial occurred 4 min 50 s after placement in the chambers, the ITI was 5 min, and rats remained in the context for 1 min following the final trial.
Stage 2: serial-order conditioning
On day 5, rats received bilateral intra-BLA infusions of either vehicle (n = 13) or D-AP5 (n = 14) ∼10 min before the session of serial-order conditioning. The details for this conditioning session were identical to those described for Experiment 1. In brief, it consisted in four exposures to the sequence S2-S1-shock. In each sequence, offset of the 30-s S2 co-occurred with onset of the 10-s S1, which then co-terminated with a 0.8 mA × 0.5 s foot shock.
Test
Rats were tested for freezing to S2 on day 7 and to S1 on day 8. These tests were identical to those described for Experiment 1.
Procedure for Experiment 3a
Context exposure
The details of context exposure on days 1–2 and context extinction on days 4, 6, and 7 were the same as in Experiment 1.
Stage 1: conditioning of S1
On day 3, rats received four 10-s presentations of S1 each of which co-terminated in a 0.8 mA × 0.5 s foot shock. The first S1 presentation occurred 5 min after placement in the chambers, the ITI was 5 min, and rats remained in the chambers for 1 min following the final shock.
Stage 2: serial-order conditioning with a trace interval between S1 and shock
On day 5, rats received bilateral intra-BLA infusions of either D-AP5 (n = 12) or vehicle (n = 12) ∼10 min before the serial-order session. This consisted in four S2-S1-[10”]-shock sequences. Each sequence consisted of a 30-s S2, which co-terminated in the onset of a 10-s S1, whose termination was followed 10 s later by foot shock. The first sequence occurred 5 min after the rats were placed in the chambers, the ITI (S1 offset to S2 onset) was 5 min, and rats were removed from the chambers 1 min after the final sequence.
Test
Rats were tested for freezing to S2 and S1 on days 7 and 8, respectively, in the manner described for Experiment 1.
These rats proceeded to Experiment 3b, which began on day 9.
Procedure for Experiment 3b
Extinction of S1 and S2
On each of days 9–11, rats received two extinction sessions, one in the morning and one in the afternoon. One session consisted in 16 S1 alone presentations and the other in eight S2 alone presentations, with the order counterbalanced across sessions.
Stage 1: reconditioning of S1
On day 12, rats received four 10-s exposures to S1 each of which co-terminated in the offset of a 0.8 mA × 0.5 s foot shock. The details of the session were those described for conditioning of S1 in the previous experiments. For each rat, the stimulus reconditioned as S1 in Experiment 3b was the same as the stimulus conditioned as S1 in Experiment 3a.
Context extinction
On day 13, rats received two exposures to the chambers in the absence of any scheduled events, each 20 min in duration, with one the morning and the other 3 h later in the afternoon.
Stage 2: serial-order conditioning
On day 14, rats again received bilateral intra-BLA infusions of either D-AP5 (n = 12) or vehicle (n = 12) before the serial-order conditioning session. The drug and vehicle groups were switched such that rats who had been infused with D-AP5 in Experiment 3a now received vehicle, and vice versa. The serial-order conditioning session consisted in four S2-S1-shock sequences where the offset of the 30-s S2 co-occurred with onset of the 10-s S1, which co-terminated with a 0.8 mA × 0.5 s foot shock. The remaining details of the session were those described previously.
Context extinction
On day 15, rats received two 20-min exposures to the chambers in the absence of any events, one in the morning and the other in the afternoon. On the morning of day 16, rats received an additional 10-min exposure to the chambers.
Test
Rats were tested for freezing to S2 and S1 on the afternoon of days 16 and 17, respectively, in the manner described previously.
Procedure for Experiment 4
Context exposure
The details of context exposure on days 1–2 and context extinction on days 4, 6, and 7 were those described previously.
Stage 1: conditioning of S1 with a 0.4-mA shock
On day 3, rats received four 10 s presentations of S1, each of which co-terminated in a 0.4 mA × 0.5 s foot shock. The first S1-shock pairing occurred 5 min after placement in the context, the ITI was 5 min, and the rats remained in the context for 1 min following the final pairing.
Stage 2: serial-order conditioning with a 0.8-mA shock
On day 5, rats received bilateral intra-BLA infusions of either D-AP5 (n = 14) or vehicle (n = 13) shortly before the serial-order conditioning session which consisted in four S2-S1-shock sequences. Each 30-s S2 terminated in the onset of a 10-s S1, which co-terminated with a 0.8 mA × 0.5 s foot shock. The remaining details of the session were those described previously.
Test
Rats were tested with S2 and S1 on days 7 and 8, respectively, in the manner described previously.
Procedure for Experiment 5a
Context exposure
The details of context exposure on days 1–2 and context extinction on days 4, 6, and 7 were those described previously.
Stage 1: conditioning of S1 with a 0.8-mA shock
On day 3, rats received four 10-s presentations of S1 each of which co-terminated in a 0.8 mA × 0.5 s foot shock. The details of the session were those described previously.
Stage 2: serial-order conditioning with a 0.3-mA shock
On day 5, rats received bilateral intra-BLA infusions of either D-AP5 (n = 12) or vehicle (n = 12) ∼10 min before the serial order conditioning session. This consisted in four S2-S1-shock sequences. Each sequence consisted in a 30-s presentation of S2, which terminated in the onset of a 10-s S1, which, in turn, co-terminated in a 0.3 mA × 0.5 s foot shock. The remaining details of this session were those described previously.
Test
Rats were tested for freezing to S2 and S1 on days 7 and 8, respectively, in the manner described previously.
These rats proceeded to Experiment 5b, which began on day 9.
Procedure for Experiment 5b
Extinction of S1 and S2
On days 9 and 10, rats received two 20 min extinction sessions each day, one with S1 and the other with S2. The extinction sessions were 3 h apart and the order in which S1 and S2 were extinguished was counterbalanced. Each session consisted in eight stimulus alone presentations (either S1 or S2) with an ITI of 2 min. On day 11, rats received an additional extinction session of S1.
Stage 1: reconditioning of S1 with a 0.3-mA shock
On day 12, rats were reconditioned with S1. This consisted in four 10-s S1 presentations each of which co-terminated in a 0.3 mA × 0.5 s foot shock. The remaining details of the session were those described previously.
Context extinction
On day 13, rats received two 20-min exposures to the chambers in the absence of any scheduled events, one in the morning and the other 3 h later in the afternoon.
Stage 2: serial-order reconditioning with a 0.3-mA shock
On day 14, rats received bilateral intra-BLA infusions of either D-AP5 (n = 11) or vehicle (n = 13) shortly before the serial-order conditioning session. The drug and vehicle groups were switched such that rats who had received D-AP5 in Experiment 5a now received vehicle, and vice versa. The serial-order conditioning session consisted in four S2-S1-shock sequences. Each sequence involved presentation of a 30-s S2, which terminated in onset of the 10-s S1, which, in turn, co-terminated in a 0.3 mA × 0.5 s foot shock. The remaining details of the session were those described previously.
Context extinction
On day 15, rats received two 20-min exposures to the chambers in the absence of any scheduled events, one in the morning and the other in the afternoon. On the morning of day 16, rats received an additional 10-min exposure to the chambers to extinguish any spontaneously recovered context-elicited freezing.
Test
Rats were tested for freezing to S2 and S1 on the afternoon of days 9 and 10, respectively, in the manner described previously.
Results
Experiment 1
Williams-Spooner et al. (2022) showed that conditioning of S2 does not require NMDAR-activation in the BLA when the shock that occurs across the S2-S1-shock sequences is that predicted by the already-conditioned S1. This was demonstrated using male rats. The aim of Experiment 1 was to replicate this finding using female rats. Two groups of rats were exposed to S1-shock pairings in stage 1 and, 48 h later, to S2-S1-shock sequences in stage 2. The groups differed with respect to the BLA infusion that they received immediately before the stage 2 conditioning session: one group was infused with D-AP5 (Group D-AP5) while the other was infused with vehicle (Group VEH). Finally, rats were tested with presentations of the S2 alone and S1 alone. Following Williams-Spooner et al., we expected that the BLA infusion of D-AP5 would have no effect on conditioning to S2 (or S1) and, hence, that rats in Groups D-AP5 and VEH would freeze equally when tested with the S2 (and S1). This result would indicate that, when the S1-shock relation is preserved across the shift from stage 1 to stage 2, conditioning to S2 occurs independently of NMDAR-activation in the BLA.
Stage 1: conditioning of S1
Figure 2B shows the mean (±SEM) levels of freezing to S1 across its conditioning in Groups D-AP5 (n = 9) and Vehicle (n = 7). Freezing to S1 increased across the four S1-shock pairings in stage 1 (F(1,14) = 156.1, Fc = 4.60, p < 0.001, η2p = 0.92, CI [2.591, 3.665]), with no significant between-group difference in the rate of this increase or in the overall levels of freezing elicited by S1 (F(1,14) values < 0.6, ps > 0.45).
A, Schematic of the procedure for Experiment 1 (top) including the protocol used to verify the efficacy of the intra-BLA D-AP5 infusions (bottom). B, Mean (±SEM) levels of freezing to S1 across S1-shock pairings in stage 1. C, Mean (±SEM) levels of freezing to S2 (left) and S1 (right) across S2-S1-shock sequences in stage 2. D, Mean (±SEM) levels of freezing across test presentations of S2 alone (left) and S1 alone (right). E, Mean (±SEM) levels of freezing across S2-shock pairings in reconditioning. F, Mean (±SEM) levels of freezing across test presentations of the re-conditioned S2 alone. Other than those with misplaced cannulas, no rats were excluded from Experiment 1a which demonstrated that acquisition of freezing to S2 can occur independently of NMDAR activation in the BLA. For Experiment 1b, we did not carry forward all the rats from Experiment 1a. Instead, we carried forward just those rats: (1) that exhibited clear extinction of freezing to S2 across its prior testing (a minimum requirement for assessing the effects of a BLA D-AP5 infusion on reacquisition of fear to the S2); and (2) for which we were confident in the patency of their cannulas. The baseline level of freezing in the context is that observed across the first 2 min of test sessions, before presentation of the S2 or S1. Significant group differences (p < 0.05) are denoted by the asterisk (*).
Stage 2: serial-order conditioning
Freezing to S2 increased across the four S2-S1-shock sequences (Fig. 2C, left) in both groups (F(1,14) = 47.4, Fc = 4.60, p < 0.001, η2p = 0.77, CI [0.858, 1.633]), while freezing to S1 (Fig. 2C, right) was stable across the four sequences (F(1,14) = 2.98, p = 0.11). There were no differences in overall freezing between the groups (Fs(1,14) < 0.5, ps > 0.49) and no group × trial interactions (Fs(1,14) < 1.1, ps > 0.31).
Test
Figure 2D shows the mean (±SEM) levels of freezing in Groups D-AP5 and VEH across test presentations of the S2 (left) and S1 (right). There were no between-group differences in overall freezing to either stimulus (Fs(1,14) < 1.4, Fc = 4.60, p > 0.26), no significant changes in freezing across trials, and no group × trial interactions (Fs(1,14) < 3.7, ps > 0.07). A closer inspection of the left panel in Figure 2D shows that, across the first four trials of testing with the S2, rats in Group D-AP5 exhibited numerically less freezing than rats in Group VEH. Given the importance of the S2 test result in the context of the study and claims regarding the involvement of NMDAR in Pavlovian fear conditioning, we conducted further analyses of freezing to S2 to determine whether there was, in fact, any significant difference between the groups across the initial trials of testing. Specifically, for each rat we calculated the mean level of freezing to S2 across the first four trials of the test session. We then compared rats in Groups D-AP5 and VEH with respect to these mean levels. This comparison generated an F(1,14) statistic of 3.89 which was well below the F critical of 4.60 (p = 0.069). We additionally compared rats in Groups D-AP5 and VEH with respect to the mean levels of freezing on the first two trials of the test session. This comparison generated an F(1,14) statistic of 3.21 which was again well below the F critical of 4.60 (p = 0.095). Thus, even when we restricted our analysis to just the portion of testing where rats in Group D-AP5 froze less than rats in Group VEH, the difference between the two groups was not statistically significant. As such, these findings replicate the principal finding of our earlier study (Williams-Spooner et al., 2022) in female rats: when foot shock is signaled by an already-conditioned S1, the acquisition of freezing to S2 occurs independently of NMDAR-activation in the BLA.
Reconditioning of S2
In order to verify the efficacy of the D-AP5 administration, S2 was extinguished and then reconditioned (through its pairings with shock) following an infusion of either D-AP5 or vehicle into the BLA. For this part of the experiment (1b), we carried forward just those rats: (1) that exhibited clear extinction of freezing to S2 across its prior testing (a minimum requirement for assessing the effects of a BLA D-AP5 infusion on reacquisition of fear to the S2); and (2) for which we were confident in the patency of their cannulas. Levels of freezing to S2 (Fig. 2E) linearly increased across the reconditioning trials (F(1,5) = 7.60, Fc = 6.61, p = 0.040, η2p = 0.60, CI [0.068, 1.953]) with no differences in overall freezing between the groups (F(1,5) = 0.24, p = 0.65) and no group × trial interaction (F(1,5) = 3.39, p > 0.12).
Test of the re-conditioned S2
Figure 2F shows the mean (±SEM) levels of freezing during the drug-free test presentations of the reconditioned S2. Rats that received the BLA infusion of D-AP5 before the S2-shock pairings froze significantly less to S2 at test than those that received vehicle (F(1,5) = 8.19, Fc = 6.61, p = 0.035, η2p = 0.62, CI [0.163, 3.039]). Levels of freezing remained constant across the test trials for both groups, with no group × trial interaction (Fs(1,5) < 0.5, ps > 0.51).
This experiment has replicated two previous findings. The first is the finding that learning to fear S2 requires activation of NMDAR in the BLA when each of its presentations co-terminate with the US (Bauer et al., 2002). The second is the finding that learning to fear S2 does not require activation of these receptors when the shock that occurs across the S2-S1-shock sequences is predicted by the already conditioned S1 (Williams-Spooner et al., 2022). Each of these findings was previously observed using male rats. Here, we show that they are also evident among female rats, though the pattern of results requires some qualifications. In Experiment 1a, rats that received the BLA infusion of D-AP5 before the session of S2-S1-shock sequences froze less than vehicle-infused controls when subsequently tested with the S2 alone; but, consistent with our hypothesis, the difference between the two groups was not significant at any point across the test session (i.e., after two, four, or eight presentations of the S2 alone). In Experiment 1b, rats that received the BLA infusion of D-AP5 before the session of S2-shock pairings froze less than vehicle-infused controls when subsequently tested with the S2 alone; and, as expected, this difference between the two groups was statistically significant. However, the rats in this part of the experiment were the same as those that had been trained in its earlier component (1a) and, hence, were not naive with respect to the context and stimuli. Nonetheless, we are confident in the reliability of both findings as the apparent between-group difference in freezing to S2 in Experiment 1a was entirely attributable to the low performance of two rats in Group D-AP5 that were outliers in the combined sample of all rats (their mean levels of freezing to S2 were greater than 2 SDs below the mean levels of freezing to S2 among the remaining rats); and the between-group difference in Experiment 1b adds to findings showing that the BLA is critical for standard Pavlovian fear conditioning in both naive and non-naive rats (Lay et al., 2018; Leidl et al., 2018). With these qualifications in mind, we take the results of Experiments 1a and 1b to suggest that at the level of the BLA, the involvement of NMDA receptors in Pavlovian fear conditioning is not dissimilar in female and male rats. When S2 is directly paired with shock, its conditioning requires activation of NMDAR in the BLA; but when S2 is shocked in compound with an already conditioned S1, its conditioning does not require activation of NMDAR in the BLA.
It is worth reiterating that, as part of our protocol, rats receive extensive context exposure/extinction. This is intended to minimize/reduce conditioning of the context and, thereby, (1) facilitate acquisition of freezing to the S1 and S2 in stages 1 and 2; and (2) permit assessment of freezing to the S2 and S1 at test unconfounded by freezing to the context. In this and all subsequent experiments, the context exposures were highly effective in each of these respects: the levels of freezing to S2 and S1 in vehicle-infused controls were robust, indicating that their conditioning was successful, whereas the level of freezing in the context alone at test (before any stimulus presentation) was typically low (always <10%). Importantly, in each of the experiments, the levels of context freezing were not affected by the BLA infusion of D-AP5: that is, in each experiment, rats that received the BLA infusion of D-AP5 before the session of serial-order conditioning in stage 2 exhibited just as much freezing to the context across the next day of context alone exposure as the vehicle-infused controls (mean freezing levels in the first and last 3 min of the context extinction sessions were 31.5% for Group VEH and 30.25% for Group D-AP5; F(1,14) = 0.03, Fc = 4.60, p = 0.87). This could indicate that context conditioning occurs independently of NMDAR activation in the BLA. However, it could also be related to the fact that the context was highly familiar by the time of the serial-order conditioning session in stage 2, or the fact that the context itself had been previously conditioned in stage 1. That is, as the study was not designed to assess the impact of the BLA D-AP5 infusion on context conditioning, we cannot offer a strong interpretation of the results in relation to context. They are, however, noted here for completeness.
Experiment 2
This experiment examined whether a change in when the shock occurs (relative to the S1) across the shift from S1-shock pairings in stage 1 to S2-S1-shock sequences in stage 2 reinstates the NMDAR-activation requirement for conditioning to the S2. Rats (males, for the reasons stated above) were first conditioned to fear S1 using a trace protocol in which offset of the S1 was followed 10 s later by foot shock (S1-[trace]-shock, stage 1). They were then conditioned to fear S2 using a serial-order protocol in which offset of the S2 was immediately followed by the S1, which now co-terminated in foot shock (S2-S1-shock, stage 2). The latter conditioning session was preceded by a BLA infusion of D-AP5 or vehicle (see Fig. 3A for the design). If the unexpected arrival of shock at S1 offset reinstates the involvement of BLA NMDAR in conditioning of the S2, rats exposed to the S2-S1-shock sequences under a BLA infusion of D-AP5 should freeze less when tested with S2 than rats exposed to these sequences under a BLA infusion of vehicle alone.
A, Schematic of Experiment 2. B, Mean (±SEM) levels of freezing across S1-[10”]-shock pairings in stage 1. C, Mean (±SEM) levels of freezing to S2 (left) and S1 (right) across S2-S1-shock sequences in stage 2. D, Mean (±SEM) levels of freezing across test presentations of S2 alone (left) and S1 alone (right). The baseline level of freezing in the context is that observed across the first 2 min of test sessions, before presentation of the S2 or S1. Significant groups differences (p < 0.05) are denoted by the asterisk (*).
Stage 1: trace conditioning of S1
Figure 3B shows the mean (±SEM) levels of freezing to S1 across its trace conditioning in Groups D-AP5 (n = 12) and Vehicle (n = 10). Freezing to S1 increased across conditioning trials in both groups (F(1,20) = 56.1, Fc = 4.35, p < 0.001, η2p = 0.74, CI [1.266, 2.243]). There was no significant between-group difference in the rate of this increase or in the overall levels of freezing elicited by S1 (Fs(1,20) < 0.1, ps > 0.75).
Stage 2: serial-order conditioning
Freezing to S2 increased across the S2-S1-shock sequences (Fig. 3C) in both groups (F(1,20) = 57.6, Fc = 4.35, p < 0.001, η2p = 0.74, CI [0.924, 1.624]). Rats in Group D-AP5 froze slightly less than those in Group Vehicle; a difference that approached but did not reach a conventional level of significance (F(1,20) = 4.04, Fc = 4.35, p = 0.058). The group × trial interaction was not significant (F(1,20) < 0.2, p > 0.66). Freezing to S1 (Fig. 3C) also increased across trials (F(1,20) = 10.1, Fc = 4.35, p = 0.005, η2p = 0.34, CI [0.213, 1.021]) with no differences in overall freezing between the groups (F(1,20) = 2.47, Fc = 4.35, p = 0.13) and no group × trial interaction (F(1,20) < 0.001, p > 0.97).
Test
Figure 3D shows the mean (±SEM) levels of freezing during the drug-free test presentations of S2 (left) and S1 (right). The BLA infusion of D-AP5 before the serial-order session impaired conditioning of S2, as rats in Group D-AP5 froze significantly less to S2 than those in Group Vehicle (F(1,20) = 9.22, Fc = 4.35, p = 0.007, d = 0.89, CI [0.275, 1.485]). The linear trend contrast was not significant nor was the group × trial interaction (Fs(1,20) < 0.38, ps > 0.54), indicating that the between-group difference in freezing persisted across the S2 alone presentations. The BLA infusion of D-AP5 also impaired conditioning of S1, as rats in Group D-AP5 froze significantly less to S1 than those in Group Vehicle (F(1,20) = 5.25, Fc = 4.35, p = 0.034, d = 0.69, CI [0.061, 1.306]). This difference in freezing also persisted across the S1 presentations, as there was no significant group × trial interaction (F(1,20) = 0.62, Fc = 4.35, p = 0.44). The linear trend contrast was also not significant (F(1,20) = 1.47, p = 0.24).
This experiment has shown that when rats are shifted from trace conditioning of S1 in stage 1 to S2-S1-shock sequences in stage 2, a BLA infusion of D-AP5 disrupts the conditioning of S2. This result contrasts with the failure of D-AP5 to disrupt conditioning of S2 in the previous experiment and in the study by Williams-Spooner et al. (2022). The protocols used to condition S2 were identical in the two experiments, but the protocols used for the prior conditioning of S1 were different. In Experiment 1 and the study by Williams-Spooner and colleagues, the S1 and shock co-terminated in both stages of training, meaning that the shock occurred as and when expected during the conditioning of S2. By contrast, in the present experiment, the S1 and shock were presented 10 s apart in stage 1 but co-terminated in stage 2, meaning that the shock occurred earlier than expected during the conditioning of S2. Thus, the pattern of results across these experiments is consistent with the proposal that a change in timing of the shock (relative to the S1) across the shift from stage 1 to stage 2 reinstated the NMDAR-activation requirement for conditioning to the S2.
However, it is important to note that the BLA infusion of D-AP5 also disrupted freezing to S1 across serial-order conditioning, in contrast to the failure of the drug to disrupt this freezing in Experiment 1 and the study by Williams-Spooner et al. (2022). These contrasting results are likely because of differences in the initial conditioning of S1 in the present and past experiments. Unlike in past experiments where freezing to S1 remained constant across serial-order conditioning, freezing to S1 increased across serial-order conditioning in the present experiment, suggesting that the shift from its trace conditioning in stage 1 to serial-order conditioning in stage 2 brought about an increase in the strength of the S1-shock association. The implications of this suggestion are addressed in Experiment 3a.
Experiment 3a
The results of the previous experiment are consistent with the proposal that prediction error engages NMDAR in the BLA for Pavlovian fear conditioning (Williams-Spooner et al., 2022) but leave open questions regarding the source of the error. That is, when rats were exposed to the S2-S1-shock sequences in stage 2 after conditioning of S1 with a 10-s trace interval between its offset and shock in stage 1, the prediction error may have resulted from: (1) the unexpected early arrival of the shock (i.e., at the offset of S1 rather than 10 s later); (2) the fact that S1 had been only weakly conditioned in stage 1 and, hence, was a poor predictor of the shock; or (3) some combination of these two sources.
The aim of Experiment 3a was to discriminate between these explanations. To do so, the procedure used in Experiment 2 was repeated but with the temporal relations between S1 and shock reversed across the two stages of training. That is, in stage 1, rats received presentations of S1 each of which co-terminated in the shock (S1-shock). Then, in stage 2, rats received a BLA infusion of D-AP5 or vehicle before the conditioning session in which each presentation of S2 was immediately followed by S1, whose offset was followed 10 s later by shock (S2-S1-[10-s trace]-shock). Thus, as in the previous experiment, S1 enters stage 2 predicting that shock will occur as well as when it will occur (see Fig. 4A for the design). However, unlike in the previous experiment, S1 is strongly conditioned and the shock occurs later than predicted by the S1. If the difference between the actual and predicted time at which shock occurs is sufficient to engage NMDAR in the BLA for conditioning to the S2, then rats exposed to the S2-S1-[10-s trace]-shock sequences under D-AP5 will freeze less when tested with S2 than control rats.
A, Schematic of Experiment 3a. B, Freezing to S1 across its pairings with shock. C, Freezing to S2 (left) and S1 (right) across the S2-S1-[10”]-shock sequences. D, Test level of freezing across blocks of S2 alone (left) and S1 alone (right) presentations. The baseline level of freezing in the context is that observed across the first 2 min of test sessions, before presentation of the S2 or S1. Significant groups differences (p < 0.05) are denoted by the asterisk (*).
Stage 1: conditioning of S1
Figure 4B shows freezing to S1 across its pairings with shock for rats in Groups D-AP5 (n = 10) and Vehicle (n = 11). The analysis confirmed that freezing increased across the pairings (F(1,19) = 99.4, Fc = 4.38, p < 0.001, η2p = 0.84, CI [1.508, 2.310]), and that there were no significant between group differences in the levels of freezing nor a significant group × trial interaction (Fs(1,19) < 0.2, ps > 0.66).
Stage 2: serial-order conditioning with a 10-s trace interval to the shock
Figure 4C shows freezing to S2 (left) and S1 (right) across the S2-S1-[10-s trace]-shock sequences. Freezing to S2 increased across the sequences in both groups (F(1,19) = 56.5, Fc = 4.38, p < 0.001, η2p = 0.75, CI [1.006, 1.782]) who did not differ in their overall levels of freezing (F(1,19) = 2.4, Fc = 4.38, p = 0.14) or in the rate of increase across the sequences (F(1,19) = 0.05, p = 0.83). The levels of freezing to S1 did not change across the sequences (F(1,19) < 0.05, p > 0.83), and there was no significant between-group differences in the overall levels of freezing (F(1,19) = 3.74, Fc = 4.38, p = 0.068) or the rate of change across the sequences (F(1,19) = 0.05, p = 0.83).
Test
Figure 4D shows the mean (+SEM) levels of freezing across test presentations of S2 (left) and S1 (right). Inspection suggests that the BLA infusion of D-AP5 impaired conditioning of S2 without affecting conditioning of S1. The statistical analysis confirmed that rats in Group D-AP5 exhibited less freezing to S2 than those in Group Vehicle (F(1,19) = 5.55, Fc = 4.38, p = 0.029, d = 0.87, CI [0.093, 1.576]). Neither the linear trend contrast nor the group × trials interaction was significant (Fs < 1, ps > 0.33), indicating that the between group differences in freezing remained constant across the test presentations. The analysis also confirmed that there were no significant between-group differences in the levels of freezing to S1 (F(1,19) = 0.43, p = 0.52), nor any change in freezing across the presentations or group × trial interaction (Fs(1,19) < 1.5, ps > 0.23).
This experiment has shown that a BLA infusion of D-AP5 disrupts conditioning to S2 when rats are shifted from S1-shock pairings in stage 1 to S2-S1-[10-s trace]-shock sequences in stage 2. Importantly the effect of the D-AP5 infusion on conditioning to S2 was observed in the absence of any effect on freezing to S1, which remained high across stage 2. The fact that S1 entered stage 2 eliciting high levels of freezing rules out an explanation of the S2 result in terms of S1 being a weak predictor of shock. Instead, it implies that the prediction error that engages BLA NMDAR for conditioning to S2 reflects more than the discrepancy between observed and expected shock. That is, we take the present findings and those of the previous experiments to mean that changes in timing of the shock (relative to the S1) across the shift from stage 1 to stage 2 reinstate the NMDAR-activation requirement for conditioning to the S2. NMDAR in the BLA are not engaged for conditioning to S2 if the shock occurs as and when expected based on the S1; but are engaged for this conditioning if the shock occurs earlier (Experiment 2) or later than predicted by the S1 (Experiment 3a).
Experiment 3b
The aim of Experiment 3b was to provide a further replication of the Williams-Spooner et al. (2022) finding that conditioning to S2 does not require activation of NMDAR in the BLA when: (1) it is shocked in compound with an already-conditioned S1; and (2) the shock is that predicted by the S1. Following the completion of Experiment 3a, the rats received repeated S2 alone and S1 alone presentations to extinguish their ability to elicit freezing responses. They were then trained using the protocol described in Experiment 1a. Briefly, rats were exposed to presentations of S1 each of which co-terminated with shock; and, 2 d later, S2-S1-shock sequences where each S2 presentation was immediately followed by S1, which continued to co-terminate with shock. Minutes before the latter conditioning session, rats received a BLA infusion of D-AP5 or vehicle. Finally, rats were tested for freezing to S2 and S1 (see Fig. 5A for the experiment design). The physical identities of the stimuli designated S1 and S2 were the same as those used for each rat in Experiment 3a. However, the rats that had been infused with vehicle before the serial-order conditioning session in Experiment 3a were now infused with D-AP5 and vice versa. Based on previous results (Experiment 1; Williams-Spooner et al., 2022), we expected that the BLA infusion of D-AP5 would fail to affect conditioning of S2 in contrast to its disruptive effect in Experiment 3a.
A, Schematic of Experiment 3b. B, Mean (±SEM) levels of freezing to S1 across its reconditioning. C, Mean (±SEM) levels of freezing to S2 (left) and S1 (right) across the S2-S1-shock sequences. D, Mean (±SEM) levels of freezing to S2 (left) and S1 (right) across blocks of two trials at test. The baseline level of freezing in the context is that observed across the first 2 min of test sessions, before presentation of the S2 or S1. Significant groups differences (p < 0.05) are denoted by the asterisk (*).
Extinction of S1 and S2
After three daily extinction sessions of S1 and S2, the mean levels of freezing on the final S1 and S2 presentations were 10.9% and 6.1%, respectively, for Group D-AP5, and 6% and 12%, respectively, for Group Vehicle.
Stage 1: reconditioning of S1
Figure 5B shows that reconditioning was successful. Freezing to S1 increased across its pairings with shock for both groups (F(1,19) = 7.14, Fc = 4.38, p = 0.015, η2p = 0.27, CI [0.159, 1.305]), with no main effect of Group nor any group × trial interaction (Fs(1,19) < 2, ps > 0.17).
Stage 2: serial-order conditioning
Figure 5C shows that reconditioning of the S2 element of the serial-order compound was also successful. Freezing to S2 increased across the S2-S1-shock sequences (F(1,19) = 14.96, Fc = 4.38, p = 0.001, η2p = 0.44, CI [0.379, 1.272]), with no between-group differences in overall freezing nor any group × trial interaction (Fs(1,19) < 0.4, ps > 0.53). Freezing to the already reconditioned S1 did not differ between the groups (F(1,19) = 0.88, Fc = 4.38, p = 0.36). It did decrease across the sequences (F(1,19) = 6.06, Fc = 4.38, p = 0.024, η2p = 0.24, CI [−0.870, −0.070]) but this decrease did not differ between the groups (F(1,19) = 0.52, p = 0.48). Inspection suggested that this decrease was because of the S1 coming to elicit escape (e.g., jumping) rather than freezing responses, presumably a consequence of the extended conditioning of S1.
Test
Figure 5D shows the mean (+SEM) levels of freezing across test presentations of S2 (left) and S1 (right). Inspection of the figure indicates that the BLA infusion of D-AP5 before the serial-order conditioning session failed to disrupt conditioning of either S2 or S1. The statistical analysis of the S2 data confirmed that there were no significant between-group differences in freezing (F(1,19) = 0.35, Fc = 4.38, p = 0.56), nor any main effect of trial or group × trial interaction (Fs(1,19) < 2.75, ps > 0.11). Similarly, there were no significant between-group differences in freezing to S1 (F(1,19) = 0.76, Fc = 4.38, p = 0.39), nor any significant decrease across the presentations (F(1,19) = 3.5, Fc = 4.38, p = 0.077) or group × trial interaction (F(1,19) < 0.26, p > 0.61). These results provide a further replication of those reported by Williams-Spooner et al. (2022) and observed in Experiment 1a: conditioning to the S2 element of the serial-order compound does not require activation of NMDAR in the BLA when the shock occurs as and when predicted by the already-conditioned S1 element. It is important to acknowledge that the rats in this experiment were not experimentally naive, having been trained and tested in Experiment 3a. This issue is addressed in a separate section at the end of the final experiment.
Experiment 4
The previous experiments have shown that conditioning of S2 does not require NMDAR-activation in the BLA when the temporal relation between S1 and shock remains the same in stages 1 and 2 (Experiments 1 and 3b) but does require this activation when the temporal relation between S1 and shock changes in stage 2 (Experiments 2 and 3a). The aim of this experiment was to determine whether a change in shock intensity in stage 2 also engages NMDAR in the BLA for conditioning to S2. Briefly, rats were exposed to S1-shock pairings in stage 1, where the intensity of the shock was 0.4 mA; and to S2-S1-shock sequences in stage 2, where the shock intensity was increased to 0.8 mA. Critically, the latter conditioning session occurred under a BLA infusion of D-AP5 or vehicle (see Fig. 6A). We hypothesized that the increase in foot shock intensity in stage 2 would generate a prediction error and, thereby, engage NMDAR in the BLA for conditioning to S2. Hence, we expected rats that received the BLA infusion of D-AP5 before the session of S2-S1-shock sequences would freeze less when tested with S2 than the vehicle-infused controls.
A, Schematic of Experiment 4. B, Mean (±SEM) levels of freezing to S1 across its pairings with a 0.4 mA × 0.5 s shock. C, Mean (±SEM) levels of freezing to S2 (left) and S1 (right) across serial-order conditioning with a 0.8 mA × 0.5 s shock. D, Mean (±SEM) test levels of freezing across blocks of two S2 alone (left) and S1 alone (right) presentations. The baseline level of freezing in the context is that observed across the first 2 min of test sessions, before presentation of the S2 or S1. Significant groups differences (p < 0.05) are denoted by the asterisk (*).
Stage 1: conditioning of S1 with a 0.4-mA shock
Figure 6B shows the levels of freezing to S1 across its pairings with the 0.4 mA foot shock for rats in Groups D-AP5 (n = 9) and Vehicle (n = 11). Freezing increased across the pairings (F(1,18) = 65.5, Fc = 4.41, p < 0.001, η2p = 0.78, CI [1.313, 2.234]), with no significant between-group differences in the overall levels of levels of freezing or significant group × trend interaction (Fs(1,18) < 0.16, ps > 0.69).
Stage 2: serial-order conditioning with a 0.8-mA shock
Figure 6C shows freezing to S2 (left) and S1 (right) across the S2-S1-shock sequences. Freezing to S2 increased across the sequences in both groups (F(1,18) = 37.3, Fc = 4.41, p < 0.001, η2p = 0.67, CI [0.905, 1.854]). The groups did not differ in the rate of increase across the sequences or their overall level of freezing (Fs(1,18) < 0.04, ps > 0.84). The levels of freezing elicited by S1 were similar in the two groups. They were also substantial and remained unchanged across the S2-S1-shock sequences. There were no significant between-group differences in the overall levels of freezing to S1, and no significant trend or group × trend interaction (Fs(1,18) < 0.25, ps > 0.62).
Test
Figure 6D shows the mean (±SEM) test levels of freezing to S2 (left) and S1 (right). Inspection of the figure suggests that the BLA infusion of D-AP5 before the session containing the S2-S1-shock sequences impaired conditioning of S2 without affecting the already conditioned S1. The statistical analyses confirmed that rats in Group D-AP5 exhibited less overall freezing to S2 than rats in Group Vehicle (F(1,18) = 7.13, Fc = 4.41, p = 0.016, d = 0.76, CI [−1.409, −0.168]). There was a significant linear trend indicating that freezing declined across S2 alone presentations (F(1,18) = 10.9, Fc = 4.41, p = 0.004, η2p = 0.38, CI [−1.336, −0.297]). However, the difference in freezing between the two groups persisted across the test trials, as evidenced by the absence of a significant group × trial interaction (F(1,18) < 0.03, p > 0.86). The analysis also confirmed that there were no significant between-group differences in freezing to S1 nor any linear trend or group × trend interaction (Fs(1,18) < 0.63, ps > 0.43).
These results show that a BLA infusion of D-AP5 disrupted conditioning to S2 when the foot shock intensity was increased from 0.4 mA in stage 1 to 0.8 mA in stage 2. Importantly, the D-AP5 infusion disrupted conditioning to S2 without affecting S1, which elicited substantial and persistent freezing across all trials of stage 2. Hence, the effect of D-AP5 on conditioning to S2 cannot be because of S1 being a poor predictor of shock. Instead, we take these findings to mean that NMDAR in the BLA are not engaged for conditioning to S2 when the foot shock intensity remains the same in stages 1 and 2 (Experiments 1 and 3b; Williams-Spooner et al., 2022) but are engaged for this conditioning when the foot shock intensity increases in stage 2. That is, the prediction error generated by the increase in shock intensity reinstated the NMDAR-activation requirement for conditioning to the S2.
Experiment 5a
The previous experiment showed that a BLA infusion of D-AP5 disrupted conditioning to S2 when the shock intensity increased from 0.4 mA in stage 1 (S1-shock pairings) to 0.8 mA in stage 2 (S2-S1-shock sequences). The present experiment examined the effect of a BLA D-AP5 infusion on conditioning to S2 when the foot shock intensity changed in the opposite direction, i.e., when it was strong in stage 1 and weak in stage 2. Here, we were concerned that the rats’ prior experience with shock would have a sensitizing effect and, hence, that the difference between the strong and weak shocks would be functionally less than in the case where we shifted rats from the weak shock in stage 1 to the strong shock in stage 2 (Experiment 4). As such, we reduced the intensity of the weak shock in Experiment 5a from 0.4 to 0.3 mA to further enhance the discrepancy between the two shocks and, thereby, the sensitivity of our tests to any effect of changing the shock intensity. Briefly, in stage 1, rats were exposed to S1 presentations each of which co-terminated with a 0.8-mA shock. In stage 2, they received a BLA infusion of D-AP5 or vehicle and were then exposed to S2-S1-shock sequences where the S1 co-terminated with a 0.3-mA shock (see Fig. 7A for the design). If the discrepancy between the actual shock (0.3 mA) and that predicted by S1 (0.8 mA) engages NMDAR in the BLA for conditioning of S2, then rats that received the D-AP5 infusion in stage 2 will freeze less when tested with S2 than vehicle-infused controls.
A, Schematic of Experiment 5a. B, Mean (±SEM) levels of freezing to S1 across its pairings with a 0.8-mA shock. C, Mean (±SEM) levels of freezing to S2 (left) and S1 (right) across serial-order conditioning with a 0.3-mA shock. D, Mean (±SEM) test levels of freezing across blocks of two S2 alone (left) and S1 alone (right) presentations. The baseline level of freezing in the context is that observed across the first 2 min of test sessions, before presentation of the S2 or S1. Significant groups differences (p < 0.05) are denoted by the asterisk (*).
Stage 1: conditioning of S1 with a 0.8-mA shock
Figure 7B shows freezing in Groups D-AP5 (n = 12) and Vehicle (n = 10) to S1 each of whose presentations co-terminated in a 0.8-mA shock. Freezing increased across the S1-shock pairings (F(1,20) = 135.0, Fc = 4.35, p < 0.001, η2p = 0.87, CI [1.986, 2.855]) with no between-group differences in overall freezing and no group × trial interaction (Fs(1,20) < 0.79, ps > 0.38).
Stage 2: serial-order conditioning with a 0.3-mA shock
Figure 7C shows freezing to S2 (left) and S1 (right) across serial-order conditioning in which each 30-s S2 co-terminated in the onset of a 10-s S1, which, in turn, co-terminated in a 0.3 mA × 0.5 s shock. Freezing to S2 increased across the S2-S1-shock sequences (F(1,20) = 66.4, Fc = 4.35, p < 0.001, η2p = 0.77, CI [1.372, 2.315]). There was a significant main effect for group with rats in Group D-AP5 freezing significantly less overall (F(1,20) = 22.4, Fc = 4.35, p < 0.001, d = 0.91, CI [0.729, 1.875]) than those in Group Vehicle. There was also a significant group × trial interaction (F(1,20) = 11.5, Fc = 4.35, p = 0.003, η2p = 0.37, CI [0.593, 2.481]), which, from inspection, was because of the greater change in levels of freezing from trial 1 to trial 2 among the vehicle-infused than the D-AP5 infused rats. Rats in Group D-AP5 also froze less overall to S1 than the Vehicle treated rats (F(1,20) = 13.9, Fc = 4.35, p = 0.001, d = 1.09, CI [0.498, 1.765]). Freezing to S1 increased in freezing across the sequences (F(1,20) = 5.49, Fc = 4.35, p = 0.03, η2p = 0.22, CI [0.052, 0.901]), but there was no significant group × trial interaction (F(1,20) = 0.90, p = 0.35), indicating that the between-group differences persisted across the S2-S1-shock sequences.
Test
Figure 7D shows the mean (±SEM) test levels of freezing across S2 alone (left) and S1 alone (right) presentations. Inspection suggests that the infusions of D-AP5 had impaired conditioning of S2 while leaving intact conditioning of S1. The statistical analysis of the S2 test data confirmed that rats in Group D-AP5 froze less than those in Group Vehicle (F(1,20) = 5.32, Fc = 4.35, p = 0.032, d = 0.74, CI [0.072, 1.432]). There was no significant change in the levels of freezing across the test and no significant group × trial interaction (Fs(1,20) < 3.04, ps > 0.09). The analysis of the S1 test data also confirmed that there were no significant between-group differences in freezing (F(1,20) = 0.18, p = 0.68), nor a significant effect of trial or group × trial interaction (Fs(1,20) < 2.1, ps > 0.16).
This experiment has shown that a BLA infusion of D-AP5 disrupted conditioning to S2 when the foot shock intensity decreased from 0.8 mA in stage 1 (S1-shock pairings) to 0.3 mA in stage 2 (S2-S1-shock sequences). As in the previous experiment, the drug disrupted conditioning to S2 in the absence of any effect on S1, which continued to elicit substantial and persistent freezing across all trials of stage 2. Hence, the effect of the infusion on conditioning to S2 cannot be because of S1 being a poor predictor of shock. Instead, we take the present and past findings to mean that NMDAR in the BLA are not engaged for conditioning to S2 when the shock intensity remains the same in stages 1 and 2 (Experiments 1 and 3b; Williams-Spooner et al., 2022) but are engaged for this conditioning when the shock intensity changes in stage 2 (Experiments 4 and 5a). That is, both positive prediction error when the shock intensity increased (Experiment 4) and negative prediction error (Experiment 5a) when the intensity decreased reinstated the NMDAR-activation requirement for conditioning to the S2.
It is worth noting that, in previous experiments, the BLA infusion of D-AP5 had an acute but modest effect on acquisition of freezing to the S2: relative to vehicle-infused controls, rats that received the D-AP5 infusion froze less to S2 when they were under the influence of the drug, but these differences were not statistically significant. By contrast, in the present experiment where the intensity of the shock used in stage 1 (0.8 mA) was reduced in stage 2 (0.35 mA), the BLA infusion of D-AP5 had a clear acute effect on freezing to both the S2 and S1. These differential effects of D-AP5 cannot be attributed to between-experiment differences in the effectiveness of the infusion per se, as rats that appeared to have been unaffected by the infusion in stage 2 (i.e., no acute effect on freezing to S2) exhibited a clear effect of the infusion when tested drug-free in stage 3. Instead, differences in the acute effects of the drug across experiments may reflect differences in what is learned about the two stimuli and context: specifically, reducing the intensity of the shock may have resulted in some inhibitory learning about the context, which then reduced the degree to which it otherwise masked what was learned about the S2 and/or S1. In principle, we could have assessed the impact of the D-AP5 infusion on acquisition of freezing to the S2 by shifting rats to a different context for training in stage 2. However, this would have raised a number of other, more serious concerns, as we have previously shown that a context shift per se alters the involvement of BLA NMDAR in acquisition of freezing to the S2 (Williams-Spooner et al., 2022). As such, where it was observed, we take the acute effect of the D-AP5 infusion on freezing to S2 to be suggestive of a disruption in its conditioning; and the results of the drug-free testing to be the clearest indicator of how the infusion actually affected conditioning to the S2.
Experiment 5b
Experiment 5b aimed to provide an additional demonstration of NMDAR independent conditioning when S2 was: (1) paired with shock in the presence of a pretrained S1; and (2) the shock occurs as expected based on the presence of the S1. As in the demonstration of this independence in Experiment 3b, the rats used for this experiment were those from the previous experiment (Experiment 3a). However, in contrast to Experiment 3b where a 0.8-mA shock was used for reconditioning of S1 and serial-order conditioning, here a 0.3-mA shock was used. In brief, following the completion of Experiment 5a, the rats underwent S1 and S2 extinction sessions. They were then exposed to presentations of the same S1, each of which co-terminated in a 0.3-mA shock. Next, they were exposed to S2-S1-shock sequences where each 30-s S2 co-terminated in the onset of a 10-s S1, which co-terminated in the 0.3 mA × 0.5 s shock. These sequences occurred under a BLA infusion of D-AP5 if rats had previously been infused with vehicle and vice versa (see Fig. 8A for the design). We predicted a similar pattern of results in Experiment 5a and the present Experiment 5b as that found for Experiments 3a and 3b. Specifically, D-AP5 infusions into the BLA would fail to affect conditioning of S2 in this experiment, which would contrast with its disruptive effect on conditioning in Experiment 5a.
A, Schematic of Experiment 5b. B, Mean (±SEM) levels of freezing to S1 during stage 1 S1-US (0.3 mA) pairings. C, Freezing to S2 (left) and S1 (right) during stage 2 S2 S1-US (0.3 mA) sequences. D, Test levels of freezing across blocks of two S2 alone (left) and S1 alone (right) presentations. The baseline level of freezing in the context is that observed across the first 2 min of test sessions, before presentation of the S2 or S1.
Extinction of S1 and S2
Following the final test in Experiment 5a, freezing to S1 and S2 was successfully extinguished across the three daily sessions: among rats in the present Group D-AP5, the mean level of freezing on the final extinction trial was 8.9% for S1 and 2.2% for S2; among those in Group Vehicle, it was 6.2% for S1 and 4.1% for S2.
Stage 1: reconditioning of S1 with a 0.3-mA shock
Figure 8B shows freezing to S1 across its pairings with the 0.3-mA shock. Freezing increased across the pairings (F(1,20) = 17.9, Fc = 4.35, p < 0.001, η2p = 0.47, CI [0.565, 1.663]), with no significant between-group differences (F(1,20) < 2.4, p > 0.14) or group × trial interaction (F(1,20) < 0.6, p > 0.44).
Stage 2: serial-order reconditioning with a 0.3-mA shock
Figure 8C shows freezing to S2 (left) and S1 (right) across the S2-S1-shock sequences. Freezing to S2 increased across the sequences in both groups (F(1,20) = 21.3, Fc = 4.35, p < 0.001, η2p = 0.52, CI [0.600, 1.592]) with no between-group differences in overall freezing and no group × trial interaction (Fs(1,20) < 0.13, ps > 0.72). The freezing elicited by S1 was substantial and remained constant across the S2-S1-shock sequences. There were no significant between-group differences or significant linear trend or group × trial interaction (Fs(1,20) < 3.54, ps > 0.075).
Test
Figure 8D shows the mean (±SEM) test levels of freezing across blocks of two S2 alone (left) and S1 alone (right) presentations. As the left panel shows, infusing D-AP5 into the BLA before serial-order reconditioning failed to impair conditioning of S2. Thus, unlike in Experiment 5a, antagonizing NMDA receptors during the reconditioning of S2 did not disrupt conditioning of S2. This observation was confirmed by the statistical analyses, which found no significant between-group differences in overall freezing (F(1,20) = 2.14, Fc = 4.35, p = 0.16), nor any main effect of trial or Group × Trial interaction (Fs(1,20) < 0.71, ps > 0.40). There were also no between-group differences in freezing to S1 at test (F(1,20) = 0.75, Fc = 4.35, p = 0.40). The levels of freezing decreased across the S1 alone presentations, indicating extinction (F(1,20) = 6.54, Fc = 4.35, p = 0.019, η2p = 0.25, CI [0.117, 1.151]), but there was no group × trial interaction (F(1,20) < 0.16, p > 0.69).
These results, along with those from Experiment 3b, provide further evidence that NMDAR activation in the BLA is not required for conditioning of the S2 element of the serial-order compound when the shock that occurs is that predicted by the already conditioned S1: either a moderately intense 0.8 mA foot shock (Experiment 3b) or a relatively mild 0.3 mA foot shock (Experiment 5b). These results obtained with non-naive rats replicate those obtained with naive rats in Experiment 1; and the results obtained with female rats in Experiment 1 replicate those obtained with male rats in the study by Williams-Spooner et al. (2022). Together, these findings speak to the generality of our claim that activation of NMDAR in the BLA is not necessary for Pavlovian fear conditioning: the claim holds true independently of whether animals have or have not had prior experience with the to-be-conditioned stimuli. Nonetheless, it should be noted that the impact of changes in shock timing and intensity on the NMDAR-activation requirement for Pavlovian conditioning was assessed in male rats only. This was done to facilitate comparisons with the prior study by Williams-Spooner et al. (2022), which used mostly male rats; but leaves open the question of whether changes in shock timing and intensity influence the involvement of NMDAR in Pavlovian fear conditioning in female rats. The available evidence suggests that the involvement of the BLA in standard Pavlovian fear conditioning protocols is not sex-dependent (Fam et al., 2023); that the involvement of NMDAR in Pavlovian fear conditioning protocols is not sex-dependent (Qureshi et al., 2023); and, finally, that conditioning to S2 can occur independently of NMDAR in the BLA in both sexes. As such, we have every reason to believe that changes in shock timing and intensity will influence female rats in the same way that changes in these parameters influence male rats, but this awaits confirmation in future research.
The effects of context extinction, order of testing and reusing animals
In each experiment, rats received context alone exposures between stages 1 and 2; and between stage 2 and testing. These context alone exposures were intended to extinguish freezing to the context and, thereby, increase the sensitivity of our tests to conditioning of the S2. As a feature of the protocol, it is possible that the context exposures may have played some role in determining whether or not conditioning of the S2 required activation of NMDAR in the BLA. However, this is unlikely for two reasons. First, the BLA infusion of D-AP5 had the same effect on conditioning to S2 among rats that differed with respect to their levels of context exposure/extinction, i.e., the infusion disrupted conditioning to S2 in Experiments 1b, 2, and 3a although rats in Experiment 1b received considerably more context exposure/extinction; and the infusion spared conditioning to S2 in Experiments 1a, 3b and 5b although rats in Experiment 1a received considerably less context exposure/extinction. Second, the BLA infusion of D-AP5 had different effects on conditioning to S2 among rats that received equivalent amounts of context exposure/extinction. That is, (1) the infusion spared conditioning to S2 in Experiment 1a but disrupted conditioning to S2 in Experiments 2, 3a, 4, and 5a although rats in each of these experiments received the same modest amount of context exposure/extinction before the critical stage of training; and (2) the infusion disrupted conditioning to S2 in Experiment 1b but spared conditioning to S2 in Experiments 3b and 5b although rats in each of these experiments received the same extensive amount of context exposure/extinction before the critical stage of training. Hence, the effects of the D-AP5 infusion on conditioning to S2 were unrelated to the level of context exposure/extinction, and we conclude that the involvement of NMDAR in conditioning to the S2 are independent of the level of exposure to the context.
In each experiment, the principal goal was to assess the impact of a BLA D-AP5 infusion on acquisition of freezing to the S2: hence, we prioritized testing of the S2 over testing of the S1. This, of course, means that the effect of the BLA D-AP5 infusion on freezing to S1 needs some qualification. That is, in tests where rats in Group D-AP5 froze less to S1 than rats in Group VEH, this might indicate some impact of the D-AP5 infusion on additional conditioning of the S1 in stage 2. Alternatively, it might indicate some impact of the prior testing with S2 on the level of freezing to the S1. Future work could discriminate between these possibilities by replaying the designs/manipulations used here and testing rats with S2 and S1 in the reverse order, i.e., the test of S1 would precede the test of S2. It should be noted, however, that the impact of the D-AP5 infusion on test levels of freezing to S1 has no bearing on our interpretation of how this infusion affects test levels of freezing to S2; and, as such, this assessment awaits future work.
Finally, it must be noted that, in Experiments 3b and 5b, the rats were not experimentally naive: they were the same as those used in Experiments 3a and 5a, respectively. As such, the absence of a D-AP5 effect in these experiments may have been because of aspects of the rats’ conditioning history that were unrelated to low prediction error across the S2-S1-shock sequences in stage 2; for example, familiarity with the context, shock and/or stimuli used as S2 and S1. This possibility is supported by others’ findings that conditioning history determines the involvement of hippocampal NMDAR in context fear conditioning: activation of these receptors is necessary for context fear conditioning in experimentally naive subjects but not among participants that had been previously conditioned to fear a similar-but-different context (Sanders and Fanselow, 2003; Hardt et al., 2009; Wang et al., 2012; Finnie et al., 2018; see also Bannerman et al., 1995; Saucier and Cain, 1995). However, there are three reasons for supposing that, although the rats were not experimentally naive, the results of Experiments 3b and 5b reflect the impact of low prediction error. First, the results of these experiments replicate those obtained in Experiment 1a and the prior study by Williams-Spooner et al. (2022), in which rats were naive, obtained from the same source, trained in the same physical context and injected with D-AP5 using the same parameters. Second, in Experiment 1b and the prior study by Williams-Spooner and colleagues, rats with a very similar conditioning history displayed a clear effect of the BLA D-AP5 infusion on acquisition of fear to an S2. Third, even previous findings in relation to hippocampal NMDAR and context fear conditioning can be explained in terms of prediction error: these receptors are necessary for context fear conditioning when rats are experimentally naive, as the shock is not predicted by features of the novel conditioning context; but are not necessary when rats have been previously shocked in a similar-but-different context, as the shock is predicted by common features of the two contexts (for a discussion of these findings in terms of prediction error, see Williams-Spooner et al., 2022). Thus, while we acknowledge that the rats in Experiments 3b and 5b were not naive and that their training in Experiments 3a and 5a likely contributed to the results, the simplest explanation for the absence of a D-AP5 effect is that the shock was predicted by the already-conditioned S1. That is, conditioning to the S2 occurred independently of NMDAR-activation in the BLA because the shock in stage 2 was predicted by the already-conditioned S1. It will, of course, be important to provide further replications of this finding in naive rats (see Williams-Spooner et al., 2022), particularly females; and under a range of different circumstances (e.g., with other types of conditioned stimuli, such as tastes and odors).
Discussion
This study examined whether prediction error engages NMDAR in the BLA for Pavlovian fear conditioning; specifically, whether predictions about an aversive US include information about its timing and intensity, and, thereby, whether errors in these predictions engage NMDAR for conditioning under circumstances where they are not otherwise required. In each experiment, rats were exposed to S1-shock pairings in stage 1 and then to S2-S1-shock sequences in stage 2 under a BLA infusion of the NMDAR antagonist, D-AP5, or vehicle. All rats were then tested with presentations of the S2 alone and S1 alone.
Experiment 1 replicated the finding by Williams-Spooner et al. (2022) that, when the shock occurs as predicted by the already-conditioned S1, formation of the new S2 fear memory does not require activation of NMDAR in the BLA: rats exposed to the S2-S1-shock sequences under D-AP5 exhibited just as much freezing to S2 at test as rats exposed to these sequences under vehicle (see also Experiments 3b, 5b). The remaining experiments then examined whether activation of these receptors is required for conditioning to S2 when the already-conditioned S1 generates predictions about the timing or intensity of shock that are different from the actual shock. We manipulated these predictions by: (1) inserting a 10-s trace interval between the S1 and shock in one or the other stage of training so that, in stage 2, the shock occurred earlier (Experiment 2) or later (Experiment 3a) than predicted by the S1; and (2) changing the shock intensity in stage 2 so that it was stronger (Experiment 4) or weaker (Experiment 5a) than predicted by the S1. In each case, the effect of a BLA D-AP5 infusion on conditioning to S2 was the same: rats that received this infusion before conditioning in stage 2 froze less when subsequently tested with the S2 than vehicle control rats. Together with the results of Experiment 1, these findings indicate that changes in the timing (relative to S1) or intensity of shock across the shift from stage 1 to stage 2 activate NMDAR in the BLA for conditioning to S2. More generally, they show that predictions about shock include information about its timing and intensity; thus, errors in these predictions engage NMDAR for Pavlovian fear conditioning under circumstances where they are not otherwise required.
How do errors in the timing and intensity of the shock US engage NMDAR in the BLA? According to a popular theory of Pavlovian fear conditioning (Maren and Fanselow, 1996; Fanselow and LeDoux, 1999; Johansen et al., 2011), the cellular changes in the BLA that encode a CS-US association in memory are triggered by activation of NMDAR, which respond to paired presentations of a CS and US. The pairing of the CS and US is important as the NMDAR complex is both ligand-gated and voltage-gated, and each stimulus unlocks a different gate: the CS unlocks the ligand gate by releasing glutamate that binds to the NMDAR complex, whereas the US unlocks the voltage gate by depolarizing BLA neurons, thereby opening calcium channels to the postsynaptic neuron. Thus, one way that prediction error might determine whether NMDAR are engaged for fear conditioning is through its effects on the voltage gate. When the US occurs unexpectedly, it depolarizes BLA neurons and unlocks the voltage gate; hence, NMDAR are engaged for conditioning. By contrast, when the US occurs as and when predicted by other stimuli that are present, it fails to depolarize BLA neurons; hence, the voltage gate remains locked and NMDAR are not engaged for conditioning (for evidence that foot-shock more effectively depolarizes LA neurons when it is unexpected compared with when it is expected, see Johansen et al., 2010).
The idea that prediction error regulates functioning of the voltage gate on NMDAR explains why a shock that arrives at an unexpected time or intensity engages NMDAR for conditioning (Experiments 2, 3a, 4, 5a), whereas a shock that arrives as and when expected does not (Experiments 1, 3b, 5b; Williams-Spooner et al., 2022). However, this idea fails to explain previous findings that, within the BLA, NMDAR-activation is required for: (1) the acquisition of second-order conditioned fear when an S2 is paired with an already-conditioned S1, i.e., the shock US is omitted (Gewirtz and Davis, 1997; Parkes and Westbrook, 2010; Holmes et al., 2013; Williams-Spooner et al., 2022); (2) the extinction of first-order and second-order conditioned fear when S1 or S2 is repeatedly presented in the absence of the shock US; and (3) the acquisition and extinction of an association between two neutral stimuli in sensory preconditioning, at least under some circumstances (Holmes et al., 2013). Thus, the engagement of NMDAR for associative formation cannot solely depend on cell depolarization triggered by a shock US that occurs at an unexpected time or intensity.
Another possibility is that prediction error influences the threshold for neurotransmission in the BLA. Specifically, we propose that when S2 precedes additional S1-shock pairings in the serial-order protocol, it engages previously potentiated pathways and, thereby, its conditioning bypasses the NMDAR-activation requirement in the BLA. However, if the shock fails to occur as or when expected (as in second-order conditioning, extinction of first-order and second-order conditioning, and in the present Experiments 2, 3a, 4, 5a), the S2 is processed in the same way as any other novel stimulus, thereby reinstating the NMDAR-activation requirement for its conditioning. While this proposal is speculative, it is supported by findings that: (1) GABAergic interneurons regulate Pavlovian fear conditioning in the BLA (Wolff et al., 2014; Krabbe et al., 2019); (2) the involvement of these interneurons is achieved through activation of calcium permeable AMPA receptors (CP-AMPAR), not NMDAR (Mahanty and Sah, 1998; Polepalli et al., 2010, 2020); and (3) blockade of these interneurons by the CP-AMPAR antagonist, NASPM, disrupts fear to S2 when rats are exposed to S2-S1-shock sequences regardless of whether S1 has been pretrained or not (Williams-Spooner et al., 2022). Accordingly, future work will examine the relationship between prediction error and neurotransmission in the BLA; and how this relationship is instantiated in the activation of NMDAR and CP-AMPAR in the BLA.
Finally, we recognize that the role of prediction error in determining the substrates of learning and memory in the BLA is not the same as its well-established role in regulating associative change as described by classic error-correction theories (Rescorla and Wagner, 1972). Instead, the role for error identified in the present experiments is consistent with recent “model-based” theories that retain error-correction as a central feature but differ from classic theories in two key respects (Courville et al., 2006; Redish et al., 2007; Gershman et al., 2010; Cochran and Cisler, 2019). First, model-based theories propose that animals subjected to Pavlovian conditioning protocols learn about the relation between the CS and US, but also about the situation in which the CS and US are paired: essentially, subjects form a mental model of their situation which includes information about context/time, the sensory features of specific events (e.g., the CS and US), as well as the relations between them. Second, such theories argue that prediction errors regulate the use/development of these mental models, and thereby, the way that subjects process new information. Thus, in contrast to classic theories which hold that error only regulates the strength of the association between a CS and US, model-based theories propose that error is calculated based on all aspects of the situation and that the size of the error determines whether new information is assimilated into an existing model or leads to the construction of a new mental model. When error is small, the existing model is retained and new information is assimilated into this model; and when error is large, new information cannot be assimilated into an existing model, but instead, is encoded as part of a new model. Thus, such theories provide a novel way of thinking about the role of NMDAR in learning and memory. When the US is unexpected or surprisingly omitted, subjects form a new mental model to account for the surprise, which requires activation of NMDAR in the BLA. By contrast, when the US occurs as expected given the context, time and stimuli that are present, new information is assimilated into an existing mental model, and this process of assimilation occurs independently of NMDAR activation in the BLA.
In summary, the acquisition of conditioned fear does not always require activation of NMDAR in the BLA. Rather, these receptors are engaged for conditioning when the US is in some way surprising. Specifically, activation of NMDAR in the BLA is not required for Pavlovian fear conditioning when the US occurs as and when expected given other stimuli that are present (Experiments 1, 3b, 5b). These receptors are, however, required for Pavlovian fear conditioning when the US occurs earlier or later than expected (Experiments 2, 3a), and when the US that occurs is stronger or weaker than expected (Experiments 4, 5a). Hence, we conclude that prediction error engages NMDAR in the BLA for Pavlovian fear conditioning.
Footnotes
This work was supported by the Australian Research Council (ARC) Discovery Grant DP2201036501 (to R.F.W.) and the ARC Future Fellowship FT190100697 (to N.M.H.).
The authors declare no competing financial interests.
- Correspondence should be addressed to Nathan M. Holmes at n.holmes{at}unsw.edu.au